Criminal AI

Criminal AI refers to artificial intelligence systems deliberately designed or misused to facilitate illegal activities such as fraud, theft, ransomware deployment, or other cybercrimes. This category matters for enterprise governance because organizations must understand emerging AI-powered threats that target their systems, data, and operations. Compliance and security teams need robust detection frameworks and incident response protocols to defend against adversaries who leverage machine learning for sophisticated attacks, social engineering, or automated exploitation at scale.

1 item

ResearchGlobal2026-08-20

Kriminal Sells Guardrail Bypass for $12.99, Voiding Vendor-Control Assumptions

ThreatDown researchers have identified a clearnet criminal AI service called Kriminal that wraps jailbreak prompts around legitimate models including xAI Grok, Anthropic Claude, Mistral, and Llama 3.3 to resell uncensored capabilities starting at $12.99 per month. The service offers exploit development, OSINT, social engineering, and unrestricted code generation through named agent personas. The finding demonstrates that provider-level safety controls can be systematically circumvented at commodity cost, directly undermining compliance programs that treat upstream guardrails as a primary control.

jailbreak Grok Claude Mistral Llama 3.3 xAI Anthropic criminal AI